The Great Forgetting: An AI arms race is consuming thousands of rare books which are sliced, scanned and pulped
Posted by freedomforall 2 days, 13 hours ago to Technology
Excerpt:
"One of the biggest problems with AI is that when it is trained on material generated by AI the output of each generation gets worse. It’s a feedback loop that degrades performance, so there is an arms race for original human generated content and models have run out of internet texts so they are consuming books. Printed material made before 2022 is now highly valued because it is free of AI slop content.
Anthropic, owned by Google, and Meta have been sued in the US for downloading millions of pirated books to train their AI models. It has come out in court documents that to reduce the copyright risk, Anthropic has launched a program called ‘Project Panama’. This is where they quietly buy books in vast lots, and slice then down the spine, scan all the pages and pulp the remains. This is called ‘destructive scanning’. Because they are the owner of the books, technically, this is not a breach of copyright (at least in the US).
Large booksellers have noticed a surge in high volume sales starting a few months ago.
Not surprisingly, internal documents show that Anthropic didn’t want this practice to be widely known as they knew it wouldn’t win them many friends."
"One of the biggest problems with AI is that when it is trained on material generated by AI the output of each generation gets worse. It’s a feedback loop that degrades performance, so there is an arms race for original human generated content and models have run out of internet texts so they are consuming books. Printed material made before 2022 is now highly valued because it is free of AI slop content.
Anthropic, owned by Google, and Meta have been sued in the US for downloading millions of pirated books to train their AI models. It has come out in court documents that to reduce the copyright risk, Anthropic has launched a program called ‘Project Panama’. This is where they quietly buy books in vast lots, and slice then down the spine, scan all the pages and pulp the remains. This is called ‘destructive scanning’. Because they are the owner of the books, technically, this is not a breach of copyright (at least in the US).
Large booksellers have noticed a surge in high volume sales starting a few months ago.
Not surprisingly, internal documents show that Anthropic didn’t want this practice to be widely known as they knew it wouldn’t win them many friends."
My curiosity regarding tech advances, from fire use to wheels to metalurgy to agriculture etc., has been geared toward the potential of the species to seek out introspection or devolve into hedonism. Still a coin toss. Perhaps something of the nature of that cute little animated movie ... Wall-E
Rand used the idea of an immortal robot having no values in The Objectivist's Ethics in 61.
Or .... Nomad; Star Trek episode, where an "accident" creates a "sterilization program" of anything "imperfect"
Imagine if you will ....... already been done somewhere. I keep following the arrows on the circle, knowing the end is near,
Quote of the day: "Bad and cheap always trumps good and expensive when a client can’t tell the difference."
https://pastpassport.com/p/socrates-w...
A human has the choice, most of the time, to express care. The exception is found in the biologically incapable; the psychopath. A computer is a created psychopath.